Introduction

We already know how to extract useful information from data frames. Various statistics tell us a lot about the data. Nevertheless, values of mean, quantiles and standard deviations are inconvenient for understanding the whole picture.

We get the most information through our eyes, therefore the skill of presenting the data visually is on of the most powerful. By creating simple visualisations, you can make initial hypotheses and understand possible relationships between variables.

Libraries

# install.packages('ggplot2')
# install.packages('patchwork')
library(tidyverse)

Working data

Today we will work with iris dataset. It id already included in R by default.

This famous (Fisher’s or Anderson’s) iris data set gives the measurements in centimeters of the variables sepal length and width and petal length and width, respectively, for 50 flowers from each of 3 species of iris. The species are Iris setosa, versicolor, and virginica

# data(iris)
head(iris)

##   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
## 1          5.1         3.5          1.4         0.2  setosa
## 2          4.9         3.0          1.4         0.2  setosa
## 3          4.7         3.2          1.3         0.2  setosa
## 4          4.6         3.1          1.5         0.2  setosa
## 5          5.0         3.6          1.4         0.2  setosa
## 6          5.4         3.9          1.7         0.4  setosa

Adding new factor

iris_df = iris 
iris_df$Sepal.Length.Category = cut(x=iris_df$Sepal.Length,
                                    breaks = c(4,6,8),
                                    labels = c('short', 'long'))
iris_df %>% head()

##   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
## 1          5.1         3.5          1.4         0.2  setosa
## 2          4.9         3.0          1.4         0.2  setosa
## 3          4.7         3.2          1.3         0.2  setosa
## 4          4.6         3.1          1.5         0.2  setosa
## 5          5.0         3.6          1.4         0.2  setosa
## 6          5.4         3.9          1.7         0.4  setosa
##   Sepal.Length.Category
## 1                 short
## 2                 short
## 3                 short
## 4                 short
## 5                 short
## 6                 short

Basic Visualization

plot()

The simplest graph is a points, each having x and y coordinates

x = 1:10
y = seq(2,20,2) ^ 2
x

##  [1]  1  2  3  4  5  6  7  8  9 10

##  [1]   4  16  36  64 100 144 196 256 324 400

We consider that the x and y coordinates at the same positions in each of the vectors correspond to one particular point. So we have points (1,2), (2,4), (3,6) and etc.

seq(from_value, to_value, by_value)- returns a vector with values from from_value to to_valuewith a step by_value

Scatter plot

plot(x, y)

Sepal.Length vs Sepal.Width

# plot()

We can make this graph prettier

plot(#,
     main = "Sepal.Length vs. Sepal.Width", # the title
     xlab = "Sepal.Length", # Label of X-axis
     ylab = "Sepal.Width", # Label of Y-axis 
     col = "blue", # color of plot
     pch = 19, # type of dots - 19 corresponds to the painted (solid) points
     cex = 1 # size of dots
     )

All plot() style parameters

All basic R graphical parameters

Adding lines

By default plot() shows scatter plot, but we can change this behavior

plot(x, y,type = 'l')

plot(x, y, type = 'o')

plot(x, y,type = 'b')

Show scatter with line width (lwd) = 3, point character (pch) = 2 and line type (lty) =3

# plot(x,y)

Terrible plot

Line graphs should sometimes be avoided, because lines connect points in the order of their position in the vectors

x = c(10,5,9,6,8,7,2,1,4,3)
y = c(1:4, 6:8,5,10,9)
x

##  [1] 10  5  9  6  8  7  2  1  4  3

##  [1]  1  2  3  4  6  7  8  5 10  9

plot(x, y, type = "b",  pch=19)

After sorting:

df = data.frame(x, y) %>% arrange(x)
plot(df$x, df$y,type = "b", pch=19)

ggplot2 doesn’t have such problem

Advanced

# basic r advanced

indexes_to_sort = order(x)
x_sorted = x[indexes_to_sort]
y_sorted = y[indexes_to_sort]
plot(x_sorted, y_sorted, type = "b", pch=19)

NB! We can’t just sort one of the vectors because we must save the correspondence of x and y coordinates between two vectors.

Bar Plots

heights = c(Roman = 190, 
            Ann = 172, 
            Charlie = 121,
            Vlad = 183,
            Sasha = 180) # named vector
heights

##   Roman     Ann Charlie    Vlad   Sasha 
##     190     172     121     183     180

barplot(heights)

Sort (sort()) values and create barplot again

# heights = 
# barplot(heights)

We can customize this graph

barplot(heights, 
        main = "Heights of people", # the title 
        xlab = "Height", # Label of X-axis 
        ylab = "Name", # Label of Y-axis 
        col = "lightblue", # color of inner part of bars
        border = "blue", # color of borders
        horiz = TRUE, # make barplot horizontal
        xlim = c(0,200) # limits of values showd on x-axis
        )

Number of iris_df species

Create table() of iris species and use it to make a barplot with color ‘violet’ Name the axis!

# number_of_species = 
# barplot()

Now we see that our data “is balanced”

Histograms

Need only one axis!

Distribution of Sepal.Length

hist(iris_df$Sepal.Length, 
     main = "Sepal.Length distribution", 
     xlab = "Sepal.length", 
     col = "lightgreen",
     breaks = 8 # number of x-axis splits for frequency calculation in each of the resulting ranges
     )

Each bar represents frequency of this particular Sepal.Length in dataset.
For example, first bar have height of 5 - it mean, that there are 5 flowers with Sepal.Length between 4 and 4.5.

We can make bars two times thinner

hist(iris_df$Sepal.Length,
     main = "Sepal.Length distribution", 
     xlab = "Sepal.length", 
     col = "lightgreen", 
     breaks = 16) # <---- changed

Create histogram of Sepal.Width colored ‘lightblue’. Name the axis!

# hist()

Boxplots

Boxplots are very informative charts. They display similar but more information than a histogram.

boxplot(iris_df$Sepal.Length, 
        # main = "Sepal.Length", 
        ylab = "",
        xlab = "Sepal.Length", 
        col = "darkviolet",
        horizontal = T)

Grouped boxplots

boxplot(iris_df$Sepal.Length ~ iris_df$Species, 
        main = "Sepal.Length",
        xlab = "Species",
        ylab = "Sepal.Length", 
        col = "darkviolet",
        horizontal = F)

Create boxplot of Sepal.Length grouped by Sepal.Length.Category. Name axis and let it be horizontal.

# boxplot()

Boxplot is good for unimodal** similar to normaldistributions, as it doesn’t show two peaks**

ggplot

ggplot2 is the most popular package for charts creations.

Philosophy

The ggplot is based on 3 things: data, aesthetics and geoms (geomertries)
Inside the geoms there are aesthetics.
Inside the aesthetics we put the variables from the data that we want to see in the plot. These will be our axes.
An axis is not just an x and y coordinate - any aesthetic, such as a fill,color,size etc. can also be an (pseudo)axis.
Each component in the graphic is added layer by layer

Intro

ggplot(data=iris_df, # data
       mapping = aes(x=Sepal.Length, y=Sepal.Width)) + #aesthetics (axes)
  geom_point()  # geom

Obligatory components to create chart:

ggplot(data=iris_df) - data: data.frame, tibble …
ggplot(..., mappings=aes(x=Sepal.Length, y=Sepal.Width)) - aesthetics, which turned into x and y axis.
+ geom_point() - geom, at least one.

Storing plots in variables

p = ggplot(data=iris_df, mapping = aes(x=Sepal.Length, y=Sepal.Width)) + geom_point()

p

Adding new aesthetics (“axes”)

Adding right to the `ggplot()`

ggplot(data=iris_df,
       mapping = aes(x=Sepal.Length, 
                     y=Sepal.Width,
                     color=Species, 
                     size = Petal.Width,
                     shape=Sepal.Length.Category)) + 
  geom_point()

Aesthetics list

color - color
fill - filling color
size - size
shape - shape of points
stroke - stroke thickness
alpha - transparency
other specific for geoms aesthetics…

Adding to variable

p + aes(color=Species, size = Petal.Width,
        shape=Sepal.Length.Category)

Geoms

Geometries define the types of graphs in the diagram.

Geoms have specific variables

iris_df %>% 
  ggplot(aes(x=Sepal.Width)) +   # <--- here
  geom_histogram(bins = 20, fill='lightblue', col='black')

Aesthetics can also be set up within the geoms

aesthetics inside ggplot() are set for all geoms
aesthetics inside geom_..() functions are set only for this geom

iris_df %>% 
  ggplot() + 
  geom_histogram(aes(x=Sepal.Width),
                 bins = 20, fill='lightblue', col='black')  # <--- here

Adding several geoms

iris_df %>% 
  ggplot(aes(x=Species, y=Sepal.Length)) + 
  geom_boxplot(aes(fill=Species), outliers = F) +
  geom_jitter(width=0.1, size=1)

Geoms can take different data!

iris_df %>% 
  ggplot(aes(x=Sepal.Length, y=Sepal.Width)) +
  
  geom_point(data=iris_df %>% filter(Sepal.Width > 3), # data (filtered iris)
             aes(color=Species)) +
  
  geom_hex(data=iris_df %>% filter(Sepal.Width < 3)) # another data (filtered iris)

Variables and aesthetics

fill, color, size, shape, stroke and alpha can be seen outside of aes(). They can also be set to a specific fixed value.

Each geom has a certain set of parameters. In aes() there are mandatory and optional parameters, and outside it there are only optional parameters

Inside the `aes()`

iris_df %>% 
  ggplot(aes(x=Sepal.Length, y=Sepal.Width)) +
  geom_point(aes(color=Species)) # <--- here

Outside the `aes()`

ggplot(data=iris_df,
       mapping = aes(x=Sepal.Length, y=Sepal.Width)) +
  geom_point(color='blue')  # <--- here

(not) Blue scatter-plot

ggplot(data=iris_df,
       mapping = aes(x=Sepal.Length, y=Sepal.Width, color='blue')) +
  geom_point()

Themes

Built-in themes

p + theme_bw()

p + theme_classic()

p + theme_void()

Custom themes (Advanced)

p = p + theme(axis.text = element_text(size = 15),
          axis.title = element_text(size = 20),
          panel.background = element_rect(fill = 'white', color='black'))
p

Setting theme for all plots

theme_custom <- theme(
    axis.text = element_text(size = 15),
    axis.title = element_text(size = 20),
    plot.title = element_text(size = 20, hjust = 0.5),
    plot.subtitle = element_text(size = 15, hjust = 0.5),
    legend.title = element_text(size = 15),
    legend.text = element_text(size = 15),
    panel.background = element_rect(fill = "white"), 
    panel.grid = element_blank()
  )

theme_set(theme_custom)

Labels

p +  labs(x='Sepal length', 
          y='Sepal width',
          
          color = 'Iris species',
          # fill =
          # size = 
          # shape = 
          # stroke =
          
          title = 'Scatter plot', 
          subtitle = 'Subtitle',
          caption = 'It is the great plot',
          tag = 'A')

## Ignoring unknown labels:
## • colour : "Iris species"

Simple functions for labeling axes: + xlab() and + ylab()

Let’s draw!

Scatter plot

Draw scatter-plot (geom_point) with x=Petal.Length, y=Petal.Width usingclassic theme

# iris_df %>% ggplot()

Histogram

Draw histogram (geom_histogram) of Petal.Length using blue color and lightblue fill andclassic theme

# iris_df %>% ggplot()

Density plot

Draw density plot (geom_density) of Petal.Length for every Species filled differently + alpha=0.8 and classic theme

# iris_df %>% ggplot()

Boxplot

Draw a horizontal box plot (geom_boxplot) of Petal.Width for each species on the y-axis with different fill. Remove legend (why?)

# iris_df %>% ggplot()

We can combine custom and built-in themes in one plot!

Deleting the legend (as it duplicates `color` information)

+ theme(legend.position = "none")

Barplots

sepal_len_stats = iris_df %>% group_by(Species) %>% 
  summarise(mean_sepal_len = mean(Sepal.Length), 
            sd_sepal_length = sd(Sepal.Length))

sepal_len_stats

## # A tibble: 3 × 3
##   Species    mean_sepal_len sd_sepal_length
##   <fct>               <dbl>           <dbl>
## 1 setosa               5.01           0.352
## 2 versicolor           5.94           0.516
## 3 virginica            6.59           0.636

`geom_col` - if heights of bars are known (continuous data)

Draw a barplot (geom_col) of mean_sepal_len for each species on the x-axis with different fill. Remove legend.

p_col = sepal_len_stats %>% 
  ggplot(aes(x=Species, y=mean_sepal_len, fill=Species)) + 
  geom_col() + 
  theme_classic() + 
  theme(legend.position = "none")

p_col

Adding `geom_errorbar` (SD)

p_col + geom_errorbar(aes(ymin=mean_sepal_len-sd_sepal_length,
                          ymax=mean_sepal_len+sd_sepal_length),
                      width=0.3)

Three different types of values are commonly used for error bars, sometimes without even specifying which one is used.

Standard deviation

Standard error

Confidence interval

We will discuss each later

`geom_bar` - if heights are unknown (counting categorical data)

Draw a barplot (geom_col) of Sepal.Length.Category (x-axis and fill).

# iris_df %>% ggplot()

`position="stack"` (by default)

iris_df %>% 
  ggplot(aes(x=Sepal.Length.Category, fill=Species)) + 
  geom_bar() + 
  theme_bw()

`position="fill"` - scaled from 0 to 1

iris_df %>% ggplot(aes(fill=Species, x=Sepal.Length.Category)) + 
  geom_bar(position = 'fill') + theme_bw()

`position="dodge"`

iris_df %>% 
  ggplot(aes(fill=Species, x=Sepal.Length.Category)) + 
  geom_bar(position = 'dodge')  + 
  theme_bw()

Something big and difficult

ggplot(iris, aes(x = Species, y = Sepal.Length)) +
  geom_violin(aes(fill = Species), alpha = 0.4) +  # Violin plots
  geom_boxplot(fill = "lightgray", alpha = 0.6, outliers = F, width=0.2) + 
  geom_jitter(aes(fill = Species), width = 0.2, alpha = 0.5, shape=21, col='black') +  
  geom_hline(aes(yintercept = mean(Sepal.Length)), col='red', linewidth=1, lty=2) +
  geom_text(aes(label='Global mean', x = 'setosa', y=mean(Sepal.Length)+0.1), col='red') +
  labs(title = "Boxplot of Sepal Length by Species with Jittered Points",
       x = "Species",
       y = "Sepal Length (cm)") +
  theme_minimal()

New geoms used

geom_violinplot() - Violin plot
geom_hline() - horizontal line
- geom_vline() - vertical line
geom_text() - just text
- geom_label()

Plot arranging

patchwork

library(patchwork)

bar_stack = iris_df %>% 
  ggplot(aes(x=Sepal.Length.Category, fill=Species)) + 
  geom_bar() + 
  theme_bw() 

bar_fill = iris_df %>% ggplot(aes(fill=Species, x=Sepal.Length.Category)) + 
  geom_bar(position = 'fill') + theme_bw()

bar_dodge = iris_df %>% 
  ggplot(aes(fill=Species, x=Sepal.Length.Category)) + 
  geom_bar(position = 'dodge')  + 
  theme_bw()

bar_stack + bar_fill

### Collect axes and legends

bar_stack + bar_fill +
  plot_layout(guides = 'collect',
              axes = 'collect')

(bar_stack | bar_fill) / bar_dodge +
  plot_layout(guides = 'collect') +
  plot_annotation(tag_levels = 'A')

Saving the plot

To save last drawn plot:

ggsave(
  filename = './bigplot.png', # specify the filename (path to file)
  device = , # png, pdf, jpeg, svg and others
  dpi = 320, # resolution
  width = 1800,
  height = 1600,
  units = "px" #in, cm, mm
)

Save “everything between”

pdf("./dodge_barplot.pdf", width = 6, height = 6)

iris_df %>% ggplot(aes(fill=Species, x=Sepal.Length.Category)) + 
  geom_bar(position = 'dodge')

dev.off()

## quartz_off_screen 
##                 2

This structure allow save into pdf all graphical outputs that between pdf() and dev.off()

ggpubr (Advanced)

# install.packages('ggpubr')
library(ggpubr)

Barplot

iris_df %>%  ggbarplot(x="Species", y="Sepal.Width",
                       fill='Species',
                       add = "mean_sd" # calculate mean and sd
                       )

Boxplot

Statisctics inside!

p_pubr = iris_df %>%  ggboxplot(x='Species', y='Sepal.Length', col='Species')

comparisons <- list( c("setosa", "versicolor"), 
                     c("setosa", "virginica"), 
                     c("virginica", "versicolor") )

p_pubr + stat_compare_means(comparisons = comparisons, 
                            label = "p.signif")+ 
  stat_compare_means(label.y = 10)

See more about ggpubr HERE

P.S

Customizing `fill` - `scale_fill_manual()`

# Advanced -------
iris_df %>% ggplot(aes(x=Species, y=Sepal.Length)) +
  stat_summary(aes(fill=Species), geom = 'bar', fun = mean) +
  geom_jitter(aes(fill=Species), shape=21, width = 0.1, 
              stroke=0.5, col='black') +
  theme_bw() + 
  labs(title = 'Barplots with jitter') +
# ----------

scale_fill_manual(
  values=c("#e63946", "#ffba49", "#457b9d"),
  labels=c('Setosa sp.', 'Versicolor sp.', 'Virginica sp.'),
)

Customizing `color` - `scale_color_manual()`

iris_df %>% ggplot(aes(x=Sepal.Length, y=Sepal.Width, col=Species)) +
  geom_point(alpha=0.6, size=4) +
  theme_classic() + 

scale_color_manual(
  values=c("#e63946", "#ffba49", "#457b9d"),
  labels=c('Setosa sp.', 'Versicolor sp.', 'Virginica sp.'),
)

Facets

`facet_wrap()`

iris_df %>% ggplot(aes(x=Petal.Length, y=Petal.Width)) +
  geom_point(alpha=0.6, size=2) +
  theme_bw() +

# ----------------------

facet_wrap(. ~ Species)

`facet_grid()`

iris_df %>% ggplot(aes(x=Petal.Length, y=Petal.Width)) +
  geom_point(alpha=0.6, size=2) +
  theme_bw() +
# ----------------------
  facet_grid(cols = vars(Species),
             rows=vars(Sepal.Length.Category))

  # facet_wrap(Sepal.Length.Category ~ Species)

Data visualization

Introduction

Libraries

Working data

Adding new factor

Basic Visualization

plot()

Scatter plot

Sepal.Length vs Sepal.Width

All plot() style parameters

All basic R graphical parameters

Adding lines

Terrible plot

Advanced

Bar Plots

Number of iris_df species

Histograms

Distribution of Sepal.Length

Boxplots

Grouped boxplots

ggplot

Philosophy

Intro

Storing plots in variables

Adding new aesthetics (“axes”)

Adding right to the ggplot()

Aesthetics list

Adding to variable

Geoms

Geoms have specific variables

Aesthetics can also be set up within the geoms

Adding several geoms

Geoms can take different data!

Variables and aesthetics

Inside the aes()

Outside the aes()

(not) Blue scatter-plot

Themes

Built-in themes

Custom themes (Advanced)

Setting theme for all plots

Labels

Let’s draw!

Scatter plot

Histogram

Density plot

Boxplot

Deleting the legend (as it duplicates color information)

Barplots

geom_col - if heights of bars are known (continuous data)

Adding geom_errorbar (SD)

geom_bar - if heights are unknown (counting categorical data)

position="stack" (by default)

position="fill" - scaled from 0 to 1

position="dodge"

Something big and difficult

New geoms used

Plot arranging

Saving the plot

To save last drawn plot:

Save “everything between”

ggpubr (Advanced)

Barplot

Boxplot

Statisctics inside!

P.S

Customizing fill - scale_fill_manual()

Customizing color - scale_color_manual()

Facets

facet_wrap()

facet_grid()

Adding right to the `ggplot()`

Inside the `aes()`

Outside the `aes()`

Deleting the legend (as it duplicates `color` information)

`geom_col` - if heights of bars are known (continuous data)

Adding `geom_errorbar` (SD)

`geom_bar` - if heights are unknown (counting categorical data)

`position="stack"` (by default)

`position="fill"` - scaled from 0 to 1

`position="dodge"`

Customizing `fill` - `scale_fill_manual()`

Customizing `color` - `scale_color_manual()`

`facet_wrap()`

`facet_grid()`